Hi Christian,
Purely from my gut, I think backporting llvm-toolchain-17 would be less work, as our own libraries's backports would remain trivial, and there is less of a mental delta as well.
It felt rather daunting with all the unknowns, but I agree it's
less work to backport llvm-toolchain-17. I'm sure an
llvm-toolchain-17 backport would have significant value to others
in the community too. After further consideration, I'm in full
agreement that is the better approach.
We must strictly adhere to the version in testing, as far as I'm aware. And I suspect that new newer our versions, the better anyway.
Ah. I see now from the documentation that this is the case [1][2]. I suppose that's a nail in the coffin for the idea of porting a slightly newer version of the HIP Runtime that still works with clang-15.
Sincerely,
Cory Bloor
[1]: https://backports.debian.org/Contribute/
[2]: https://wiki.debian.org/BuildingFormalBackports